Download An Optical Pulse Generator from a Sinusoidal Optical Signal Using

An Optical Pulse Generator from a Sinusoidal Optical
Signal Using Sagnac Loop – Self-Sampling
Ricardo M. Ribeiro, Vinicius N. H. Silva, Andrés P. L.
Barbero and Murilo B. Carvalho
Departamento de Engenharia de Telecomunicações
Universidade Federal Fluminense
24.210-240, Niterói, RJ - Brasil
[email protected]
Abstract—It is experimentally shown an original optical pulse
generator able to produce pulses at MHz rate with few ns
timewidth at 1/3 of input period. It is based on the self-switched
Semiconductor Laser Amplifier Loop Mirror (SLALOM)
configuration. The input is a sinusoidal signal in the C-band and
the output is amplified pulses. Numerical simulations are also
carried out and shown that GHz & ps pulse generation is
possible. Besides their usefulness as an optical pulse generator,
application in self-sampling is also discussed.
Keywords—laser; sampling; ultrafast-optics; optical signal
processing; Sagnac interferometer; semiconductor optical amplifier
I.
INTRODUCTION
Many techniques have been described to generate a
solitonic or non-solitonic optical pulses such as electro-optic
modulation [1], gain-switching [2], mode-locking [3] and
interferometric [4] schemes. In the latter, the light is phasemodulated before it traverses a Mach-Zehnder interferometer.
Short optical pulses are largely used on high-capacity fibreoptic digital communications [5], but they have other
applications, e.g. time-response characterization of ultra-fast
devices and systems [6], switching of short electrical pulses
[7] and optical signal processing [8].
Stable optical pulses as those generated by mode-locked
lasers, are crucial to carry out GSa/s rate optical sampling [8].
The authors have shown an optical circuit design and numeric
simulations [9] of a real-time optical sampler based on the
Sagnac loop using a semiconductor optical amplifier (SOA) as
the nonlinear element. The device uses a train of short pulses
to sample an analogue optical signal, e.g. a Radio-over-Fibre
(RoF) signal. By using polarisation controllers inserted in the
loop, the device was firstly adjusted to be 100% reflective.
However, it was also shown [9] that the loop length should be
carefully adjusted to almost contain an integer number of
analogue signal periods. After such settings, the sampling
pulses were released and sampled (and compressed) pulses
were transmitted [9].
In this paper, we describe for the first time, in the best of
our knowledge, experimental results on the use of an SLALOM
€
configuration [10] in the self-switched regime, to generate an
118 MHz pulse train from an incoming sinusoidal optical
Ricardo M. Ribeiro thanks Foundation Capes/MEC-Brasil for the
financial support through a Post-Doctoral fellowship under the grant BEX
9096/11-6.
Frédéric Lucarz and Bruno Fracasso
Département d’Óptique, Télécom Bretagne
Technopôle Brest-Iroise-CS83818-29238, Brest, France
[email protected]
signal around 1550 nm wavelength. The latter may be an
analogue RoF carrying or not digital modulation on the RF
carrier. Computer simulations for GHz & ps pulse generation
on VPI Transmission MakerTM software are presented. The
time compression of generated pulses and their possible use to
perform self-sampling are also discussed.
II.
THE EXPERIMENTAL SET-UP
Figure 1 depicts the experimental set-up based on a Sagnac
loop using an SOA. The circuit operates on self-switching
regime where the self-phase modulation (SPM) occurs inside
the SOA. The circuit differs from those presented in [9] as to
be an o-Sampler in that no sampling (control) pulses are used
here.
A sinusoidal optical signal with PINPUT = 344 µW power is
launched into the circuit. It is generated by direct modulation
of a distributed feedback laser diode emitting 1550 nm
wavelength, driven with 25 mA bias current and + 1 dBm
Radio-Frequency (RF) voltage signal from a signal generator.
The loop is assembled by using one fibre-coupler (FC)
featuring a measured coupling ratio of ~70/30. A polarisation
controller (PC) is inserted in the loop in order to perform the
polarisation alignment of the interfering light beams in the
coupler junction. An SOA is asymmetrically inserted in the
loop. The used SOA features 21-dB maximum gain, +6dBm
saturated power and 45-nm bandwidth around 1552.2 nm
wavelength. The bias current was set at 200 mA. The length
L1 + L2 (L1 < L2) of the fibre loop is kept constant. The input
frequency fINPUT of the sinusoidal probe signal is varied
around fRES =1/TRES corresponding to an analogue signal
period TRES given by (1).
L1 + L 2 = m
c
TRES
n eff
(1)
In (1), m is an integer number, c = 3x108 m/s and neff is the
effective refractive index of the fibre. The output signal, that
eventually may be a pulse train, is collected from the
transmission arm (PT). A tuneable bandpass filter (BPF)
operating in the 1530-1561 nm wavelength range with 1.5 nm
optical bandwidth centred at λ1 = 1550 nm, was used in the
output arm of the circuit to reduce the Amplified Spontaneous
Emission (ASE) noise from the SOA. The output optical
signal is detected by an 8 GHz bandwidth photodetector. The
resulting electrical signal can be displayed on an oscilloscope.
(c)
(d)
Fig. 1. The experimental fibre-optic circuit of the pulse generator where L1 +
L2 = Sagnac loop length.
III.
RESULTS
Figures 2a-2f show the time-domain waveforms of the
output signals from the circuit of Figure 1 when the fINPUT is
varied around the fRES = 103 MHz. The latter is calculated
from (1) using L1 = 401.0 cm, L2 = 597.5 cm and neff = 1.46.
An integer number of analogue cycles should match with L1
and L2 simultaneously. A simple calculation leads to fRES1 =
103 MHz, fRES2 = 1.03 GHz and so on, remembering that an
integer number of analogue cycles automatically should also
be fitted along the L1 + L2 loop length.
Figure 2a shows that for fINPUT = 98 MHz, the output signal
is always sinusoidal shaped despite the PC adjustment. Only
the amplitude varies when the PC is adjusted.
As is shown in Figure 2b, after increasing the input
frequency as fINPUT = 107 MHz, the output signal is no longer
sinusoidal. It starts to be distorted. Figure 2c shows that for
fINPUT = 117 MHz there is a trend to double the frequency of
the output signal.
Figure 2d shows that for fINPUT = 118.3 MHz and after a
careful adjustment of PC, it turns out possible to generate
optical pulses with ~ 8.4 ns period that exactly corresponds to
the original 118.3 MHz input frequency. The pulse timewidth
is ~ 2.6 ns, i.e. slightly smaller than 1/3 of the
(e)
(f)
Fig. 2. Output (PT) for input (PINPUT) sinusoidal signal at fINPUT = (a) 98 MHz,
(b) 107 MHz, (c) 117 MHz, (d) 118.3 MHz, (e) 120 MHz and (f) 141 MHz.
The scale of the horizontal axis is 5 ns/div.
analogue signal period. This result may be compared with
those similar as obtained in [1] by using a Mach-Zehnder
modulator operating in the nonlinear region of the
interferometer transfer function, but using an electrical input
signal and [3] using a different set-up. It should be observed
that 118.3 MHz does not correspond to the fRES. The
generation of pulses was also observed to be quite sensitive to
the launched frequency fINPUT and to the polarisation setting
through the PC adjustment.
Figure 2e shows that from fINPUT = 120 MHz is no longer
possible to generate pulses. By increasing further the launched
frequency, the output signal goes back to its sinusoidal shape
as is shown in Figure 2f for fINPUT = 141 MHz.
Because the fibre-coupler features ~ 70/30 coupling ratio, it
was not possible to set the Sagnac loop as to be 100%
reflective (or transmissive). Therefore, it was not possible to
erase the transmitted signal (PT).
(a)
(b)
Although the results were reproducible, instabilities were
observed mainly when pulses were generated as displayed in
Figure 2d. The reason is firstly because the pulses can only be
generated for a particular paddles adjustment of the PC.
Secondly, the used PC does not present high mechanical
stability. Thirdly, connectors were used in the optical circuit.
The use of non-mechanical PC [11] and fibre splices instead
of connectors may reduce the observed instabilities.
€
IV. DISCUSSIONS
Originally [9], an optical sampler was designed to operate
after adjustments of three optical delay lines. The first delay
line denoted Delay #0 is out of the loop and is used to
temporally align the sampling pulses with the analogue signal
to be sampled. The Delay #0 it is not used here. The second
and third delay lines denoted Delay #1 and Delay #2,
respectively, are included in the circuit design of Figure 3 used
for numerical simulations. By setting the PC, the device is also
phase biased to be 100% reflective when there are no
sampling pulses [9].
Now, the Delay #1 and Delay #2 lines are both inserted in
the loop. They allow a fine mismatch of L1 and L2 from the
exact fit of an integer number of analogue signal periods
TINPUT, i.e. L1 = n(c/neff)TINPUT ± ΔL1 and L2 = k(c/neff)TINPUT 
ΔL2, where n and k are integer numbers, and ΔL1 and ΔL2 are
the amount of additional optical delay. In the present model, a
symmetric adjustment of delay lines is assumed, i.e. ΔL1 = ΔL2
= ΔL so that the loop length is kept unaltered. Therefore, the
symmetric adjustment of the Delay #1 and Delay #2 lines
automatically ensures a fine placement deviation “x” of the
SOA in reference of the midpoint of the loop (see Figure 3).
3 dB ratio coupling. The start and end of the analogue signal
cycle is totally reflected back, i.e. PT = 0 mW according (2).
When PINPUT(t) is increased, a differential phase-shift
increased also due to the SPM effect that takes place into the
SOA. The SPM causes spectral enlargement and distortion of
input signal and is behind the mechanism of pulse generation.
Since the aforementioned ± ΔL setting of the loop delay lines
is carried out, the Δφ phase-shift is imprinted in different parts
of the CW and CCW analogue waveforms. By symmetrically
adjusting the delay lines, it opens a switching window and the
output signal PT may be generated as pulses. When the input
power increases (and decreases), the increase (decrease) of
PT(t) cycle is not linear thus leading the compression of the
cycles thereby generating pulses. It implies that no complete
switching is observed, i.e. the high power parts of the cycle
switch rather than its wings for pulses [14] and presently for
sinusoidal modulation.
Another possible explanation may be outlined. It is shown in
[3] an experimental set-up where actively mode-locked pulses
were generated from a SOA-fibre ring laser excited with 10
GHz modulated optical signal. Therefore, present
experimental set-up seems to generate non-dispersion
compensated mode-locked pulses since the L1 and L2 lengths
are properly adjusted. Both CW and CCW analogue signals
modulates the SOA gain.
The VPI Transmission MakerTM software is used here to
carry out some numerical simulations. The software solves the
nonlinear Schrödinger equation (NLSE) taking into account
the SPM effect in the bulk and isotropic SOA with low
feedback as an active (amplifier) nonlinear waveguide. It uses
the Transmission Line Model (TLM) to compute the
differential nonlinear phase-shift induced by a change in
carrier density in the semiconductor. As a result, the switching
is driven by the own input signals due to the SPM effect
occurring into the SOA.
Fig. 3. The fibre-optic circuit design of the pulse generator as used for
numerical simulations using the VPI Transmission MakerTM software, where
L1 + L2 = Sagnac loop length.
The interferometric transfer function of the transmitted (PT)
signal of an ideal Sagnac interferometer is given by (2) after
assuming all polarisations aligned [12,13].
PT (t ) = PINPUT (t )
1
1− cos(γL SOAPINPUT (t ))
2
[
]
(2)
In (2), Δφ(t) = γLSOAPINPUT(t) represents the SPM-generated
differential phase-shift between the clockwise (CW) and
counter-clockwise (CCW) analogue signals, respectively. The
Δφ(t) depends on the optical nonlinear strength γ and the
length LSOA of the SOA inserted in the loop, respectively, and
the input power PINPUT(t) of the analogue signal. The SOA
inserted in the loop is an active element and the output signal
generally will experiment an optical gain, i.e. PT > PINPUT.
When a single cycle of PINPUT(t) starts and ends, the power
magnitude is very low (Δφ ≈ 0 rad) and the SLALOM acts as
an almost perfect mirror since the modelled fibre-coupler is of
The circuit of Figure 3 was firstly probed with a fINPUT =
2.5 GHz (400 ps period) sinusoidal carrier frequency analogue
signal at PINPUT = 344 µW power and 1550 nm wavelength.
The loop length was set to be L1 + L2 ∝ 1.6 ns or 4 integer
cycles of the analogue signal, where L1 and L2 correspond to
0.4 ns and 1.2 ns, respectively. Because of limited amount of
available SOA’s data, it is here assumed LSOA = 500 µm,
linewidth enhancement factor = 6.0, null linear recombination,
bimolecular recombination = 1.0 x 10-16 m3/s, Auger
recombination = 1.3 x 10-41 m6/s and initial carrier density =
1.0 x 1024 1/m3. The bias current was set at IBIAS = 130 mA.
Figure 4a and 4b shows the plot of the calculated PT(t) for
fINPUT = 2.5 GHz when 50 ps of delay (detuning) is
symmetrically adjusted in both delay lines for 65/35 and 50/50
(3 dB fibre-coupler) coupling ratio, respectively. The relative
optical phase-bias between the CW and CCW signals must be
carefully adjusted through PC to optimise the output from the
circuit thus shaping the pulses. It is clearly the generation of ~
130 ps timewidth pulses, i.e. around 1/3 of the period. Figure
4a shows that the calculated pulse shape after assuming 65/35
coupling ratio reasonable fits the experimental measurement
shown by Figure 1d. However, when the coupling ratio is
assumed to be 50/50, more sharply pulses are calculated as is
shown by Figure 4b. Therefore, the interferometer visibility
contributes to increase the extinction ratio of the device.
Simulations also have shown that it is not possible to generate
pulses when the SOA is removed from the loop, but only
sinusoidal signals exit.
Simulations at fINPUT = 2.5 GHz shows that when an
“optical DC offset” of POFFSET ≤120 µW is superimposed with
400 µW sinusoidal amplitude (≤ 120/400 = 30%), the optical
DC component is blocked by the circuit of Figure 3. This
means that the pulses remain to be generated. For POFFSET >
120 µW the pulses start to turn out to be sinusoidal in shape.
This optical blocking was also observed in the experiments.
When the SOA is placed at the middle of the loop (x = 0), it
is not possible to generate pulses. However, after slightly
detuning of delay lines (x ≠ 0), the optical pulses can again be
generated.
Figure 5a and 5b show the plot of the calculated PT(t) at
fINPUT = 2.5 GHz when 0 ps and 100 ps delay (detuning) is
symmetrically adjusted in both delay lines for 50/50 coupling
ratio, respectively. Sinusoidal-like outputs are calculated for 0
ps and 100 ps delays despite the PC adjustment. A clearly
distorted sinusoidal signal is shown in Figure 5b for 0 ps
delay.
Now, the 0 ps symmetrical delays are kept unaltered and the
input frequency fINPUT is varied. Figures 6a-6c show plots for
fINPUT = 1.8, 2.3 and 2.7 GHz, respectively, after setting 50/50
coupling ratio.
(b)
Fig. 4. The calculated PT(t) for 50 ps symmetrical delay (350/1250 ps) at
fINPUT = 2.5 GHz for (a) 65/35 and (b) 50/50 coupling ratio.
(a)
From the simulation results displayed by the plots of
Figures 4-6, it becomes possible to conclude that the optical
delay |Δτ| adjustment has analogous effect of variation of input
frequency fINPUT. The experiments were carried out by varying
the fINPUT merely due to the convenience, i.e. the optical delay
lines were not available.
(b)
Fig. 5. The calculated PT(t) for (a) 0 ps (400/1200 ps) and (b) 100 ps
(300/1300 ps) symmetrical delay at fINPUT = 2.5 GHz and 50/50 coupling ratio.
(a)
(a)
(b)
Fig. 7. The calculated timewidth of the compressed pulses after propagation
along a DCF coil with PINPUT = 344 µW.
By increasing the PINPUT it turn out possible to obtain
further temporal compression. Figure 8 shows the calculated
time-shape of the compressed pulses by setting PINPUT = 1376
µW using the same 10 km length DCF coil now leading pulses
with ~ 75 ps timewidth, but presenting pedestals
(background).
(c)
Fig. 6. The calculated PT(t) at 0 ps symmetrical delays (400/1200 ps) and
50/50 coupling ratio for fINPUT = (a) 1.8 GHz, (b) 2.3 GHz and (c) 2.7 GHz. In
the latter, a 122.6 ps pulse timewidth was calculated.
V.
SELF-SAMPLING
Analogue-to-digital converters (ADCs) require the
sampling of the signal to be digitised. Even an electronic
ADC-quantizer may take the advantages of a real time optical
sampling [8,15] because of their high stability. Most of optical
sampling circuits use very short pulses generated from a
mode-locked laser that usually features few femtoseconds of
time-jitter [8,12,16].
Simulations from the circuit of Figure 3 were carried out for
fINPUT = 2.5 GHz, 50/50 coupling-ratio and 50 ps symmetrical
delays leading to generation of ~130 ps timewidth pulses as
shown in Figure 4b. Numerical simulations also show the
possibility of pulse compression. Figure 7 shows a temporal
compression at -3.1 ps/km rate by using a Dispersion
Compensating Fibre (DCF) featuring D = -160 ps/nm.km and
0.25 dB/km dispersion and attenuation coefficients,
respectively. It is assumed that the DCF coil is placed outside
the loop, i.e. the transmitted pulses traverse the DCF length
before photo-detection. The shape of the compressed pulses
were calculated to be quite similar of those presented in Figure
4b.
Figure 7 shows a slightly compression from ~130 ps to
~100 ps by using 10 km length of DCF coil. The spectral
bandwidth of the BPF was set to be ΔfFILTER = 100 GHz. After
repeating the calculation with ΔfFILTER = 400 GHz and LFIBRE =
1 km, the output pulses remain their timewidth around 117 ps.
Both experiments and simulations used a DFB laser diode that
may be assumed to present a null linewidth spectrum for
practical purposes.
Other more suited to compact systems are available to
obtain pulse compression, e.g. chirped Fibre Bragg Grating
(FBG) [17] or a Nonlinear Optical Loop Mirror (NOLM) [14].
The latter may be also useful to suppress the pulse pedestals
[18].
The compression generates a pulse train with a reduced
duty-cycle. By using one or more serially connected split-anddelay optical circuits [12,19], it turns out possible to produce
temporally interleaved pulses thus increasing the rate of the
output pulse train by NfINPUT, where N is an integer number.
Since N ≥ 2, the train may be used to sample the original
analogue waveform under the Nyquist theorem. Present
calculations for 2.5 GHz (400 ps period) show that at least N ~
5 are possible once the derived compressed pulses at ~75 ps
(without pedestals) are used to sample the analogue signal.
The scheme described in [9] may be used for real-time
self-sampling, but the wavelength should be converted or the
polarisation between the signal to be sampled and the
sampling pulses should be set at 90o angle each other.
Therefore, (amplified) sampling pulses may be derived from
the analogue signal to be sampled. Since an analogue RoF
(optical) periodic signal is already available, two advantages
in the use of self-sampling for digitising RoF or an opticalADC are pointed out here:
i) It does not require the use of a local mode-locked laser.
ii) Similarly to digital signals, analogue signals also presents
jitter. Considering the pulse generation mechanism, we are
lead to believe that the time-jitter present in the input signal
should be transferred to the derived sampling pulses. In that
case, the sampling pulses would be naturally synchronized to
the input. It should be noted however, that further
investigations are required to validate this assumption.
Additional measurements are also needed to assess the input to
output amplitude relationship of the self-sampling process.
This is essential to quantify the linearity, an important
parameter of ADCs.
his Post-Doctoral study during which the present work was
carried out. The authors also thank Profs. Ammar Sharaiha
and Tierry Rampone of ENIB (Plouzané, France) and INCTFotonicom/CNPq (Brasil).
REFERENCES
[1]
[2]
Fig. 8. The calculated time-shape of the compressed pulses after propagation
along 10 km length of DCF coil with PINPUT = 1376 µW.
VI.
[3]
CONCLUSIONS
In this paper, it is described a SLALOM-based device able
to generate optical pulses with timewidth ~1/3 of the period of
an analogue sinusoidal input optical signal. Neither
microwave circuit nor input electrical signal is required. The
pulses are generated when the period of the input sinusoidal
signal is slightly detuned from the fundamental period that can
be adjusted in the loop length and when the optical phase is
carefully biased by using the polarisation controller. In this
paper, the experiments were constrained to only ~100 MHz
input analogue signal mainly to demonstrate the proof-ofprinciple of the proposed optical generator.
[4]
[5]
[6]
[7]
[8]
Numerical simulations in the (1.8-2.7) GHz frequency
range have shown that it is equally possible to generate pulses
by:
[9]
i) Keeping unchanged the input period (or fINPUT) matched to
be an integer fraction of the loop length, and slightly adjusting
the symmetrical delays (detuning) of the SOA relative to the
fibre-junction.
ii) Keeping unchanged the L1 and L2 and varying the fINPUT
around fRES.
In both cases, one should also carefully to adjust the optical
phase-bias by using the controller PC.
[10]
Although much work remains to be done, the possibility of
time compression of the generated pulses and their use to
implement a scheme for self-sampling is proposed and
discussed. The sampling pulses may be extracted from the
analogue signal to be sampled and the time-jitter of both might
be automatically synchronized. This opens the possibility to
eliminate the use of local mode-locked laser in an opticalsampler and the availability of sampling pulses free of relative
jitter with the analogue signal to be sampled. Since the modelocking may be the mechanism behind pulse generation, the
stability issue should also be addressed.
[14]
The proposed and tested device is potentially optically
integrable, requires ~mW optical power amplitude for the
input analogue signal and outputs amplified pulses.
[18]
[11]
[12]
[13]
[15]
[16]
[17]
[19]
ACKNOWLEDGMENT
Ricardo M. Ribeiro thanks the team at Département d’Optique
of Telecom Bretagne and the CapilRTM platform for hosting
J. J. Veselka and S. K. Korotky, Pulse Generation for Soliton Systems
Using Lithium Niobate Modulators, IEEE Journal of Selected Topics in
Quantum Electronics, vol. 2, No 2, 1996, pp. 300-310.
H. Sayinc et al, Gain switched laser diode based all-fiber laser source
emitting simultaneously at 8 different wavelengths in the NIR region,
Advanced Solid-State Photonics (OSA/ASSP 2011), Istanbul, Turkey,
February 13-16, paper ATuB7, 2011.
J. W. D. Chi, A. Fernandez and C. Lu, Properties of Mode-Locked
Optical Pulses in a Dispersion-Managed Fiber-Ring Laser Using
Semiconductor Optical Amplifier as Active Device, IEEE Journal of
Quantum Electronics, Vol. 49, No. 1, 2013, pp. 80-88.
X. Wei, J. Leuthold and L. Zhang, Delay-interferometer-based optical
pulse generator, Optical Fiber Communication (OFC), February 22,
WL6, Los Angeles, CA, USA, 2004.
G. P. Agrawal, Fiber-Optic Communication Systems, 4th edition, Wiley,
2010.
A. J. Zilkie et al, Characterization of the ultrafast carrier dynamics of
an InAs/InGaAsP quantum dot semiconductor optical amplifier
operating at 1.55 µm, Proceeding of SPIE, Vol. 5971, 2005, 59710G.
T. Motet, J. Nees, S. Williamson and G. Mourou, 1.4 ps rise-time highvoltage photoconductive switching, Applied Physics Letters, Vol. 59,
No.12, 1991, pp. 1455-1457.
A. Khilo et al, Photonic ADC: overcoming the bottleneck of electronic
jitter, Optics Express, Vol. 20, No. 4, 2012, pp. 4454-4469.
R. M. Ribeiro, F. Lucarz and B. Fracasso, An All-Optical Sampler for
Digitising Radio-over-Fibre Transceivers, 18th European Conference on
Network & Optical Communications (NOC 2013), July 10-12, Graz,
Austria, 2013, pp. 27-34.
M. Eiselt, Optical Loop Mirror with Semiconductor Laser Amplifier,
Electronics Letters, Vol. 28, No. 16, 1992, pp. 1505-1507.
Y. Zhang et al, Complete polarization controller based on magnetooptic crystals and fixed quarter wave plates, Optics Express, Vol. 14,
No.8, 2006, pp. 3484-3490.
Y. Miyoshi et al, All-Optical Analog-to-Digital Conversion Using Splitand-Delay Technique, IEEE Journal of Lightwave Technology, Vol. 25,
No. 6, 2007, pp. 1339-1347.
G. P. Agrawal, Applications in Nonlinear Fiber Optics, Academic Press,
USA, 2001, pp. 125.
N. J. Doran, D. S. Forrester and B. K. Nayar, Experimental investigation
of all-optical switching in fibre loop mirror device, Electronics Letters,
Vol. 25, No. 4, 1989, pp. 267-269.
G. C. Valley, Photonic analog-to-digital converters, Optics Express,
Vol. 15, No. 5, 2007, pp. 1955-1982.
W. Ng, R. Stephens, D. Persechini and K. V. Reddy, Ultra-Low Jitter
Mode-Locking of Er-Fiber Laser at 10 GHz and its Application in
Photonic Analog-to-Digital Conversion, International Topical Meeting
on Microwave Photonics (MWP2000), Oxford, UK, September 11-13,
WE4.4, 2000, pp. 251-254.
A. Taketomi et al, Compression of picosecond pulses with a chirped
volume Bragg grating, Photonic Global Conference (PGC21012),
Singapore, December 13-16, 2012, pp. 1-3.
M. D. Pelusi, Y. Matsui and A. Suzuki, Pedestal Suppression from
Compressed Femtosecond Pulses Using a Nonlinear Fiber Loop Mirror,
IEEE Journal of Quantum Electronics, Vol. 35, No. 6, 1999, pp. 867874.
R. M. Ribeiro, F. Lucarz and B. Fracasso, Proposal and Design of an
All-Optical Encoder for Digitising Radio-over-Fibre Transceivers, 18th
European Conference on Network & Optical Communications (NOC
2013), July 10-12, Graz, Austria, 2013, 35-42.